地方政府债务风险非线性仿真预警系统的构建——基于粗糙集-BP神经网络方法集成的研究

在设计地方政府债务风险预警指标体系框架的基础上,吸收了粗糙集和BP神经网络等人工智能方法在数据处理上的优点,构建出基于粗糙集-BP神经网络集成的地方政府债务风险非线性仿真预警系统。选取我国2007～2009年东、中、西部地区9个县的27个样本数据,运用该非线性仿真预警系统进行了地方政府债务风险预警实证分析。研究结果表明,大部分样本地区的债务风险都处于"中警"及以上状态,地方政府债务风险普遍较高,同时,样本地区债务风险综合评价值是不断提高的,说明近年来我国地方政府债务风险呈现出不断上升的趋势。与单纯的BP神经网络仿真预警系统相比,该仿真预警系统不仅降低了BP神经网络的复杂性,节省了训练时间,而且具有更好的预警准确性和推广应用价值。

Research on the Nonlinear Simulation Early-Warning System of the Local Government Debt Risk——Based on Rough Sets and BP Neural Network Integration

Firstly,the paper designs the early-warning indicator system of the local government debt risk.Then,the paper integrates rough sets and BP neural network to build the nonlinear simulation early-warning system of the local government debt risk.Selecting china’s eastern,central and western regions 27 samples to carry out the empirical research in the period 2007-2009,the results show that Most of the debt risks in sample regions show the state of "middle warning degree" or above the state,the local government debt risk is totally high.Meanwhile,in the period 2007 to 2009,the comprehensive evaluation of the risk in all sample regions is rising,which also illustrates the local government debt risk in our country are showing a rising trend in recent years.Regarding the simulation effect,compared to the system of simple BP neural network,the RS-BP neural network system not only reduces the complexity of BP neural network,saving training time,but also has better early-warning accuracy and application value.